Inverse FFT on matrix

[Arturs Bekasovs]

While in MantidPlot, I create a new matrix (NOT MatrixWorkspace), resize it to be of size 3x3 and fill with some values, e.g. 1-9. Then I open the Analysis menu, and select Inverse FFT. It crashes the MantidPlot. Happens if you run the FFT... with Inverse option as well.

Observed in 2.5.1960 on Ubuntu 12.04 64bit.

[Andrei Savici]

Works for me in Ubuntu 13.10 and RHEL 6 (Mantid 3.0). I don't see InverseFFT, but works for FFT and select Inverse

[Michael Reuter]

I tested using the latest master on all platforms and here's what I found:

	Inverse FFT	FFT with Inverse Opt
RHEL6	Works	Works
Windows 7	Crashes	Crashes
OS X 10.8	Crashes	Crashes
Ubuntu 12.04	Crashes	Works

Looks like this does not work reliably.

[Dan Nixon]

The Analysis>FFT... dialogue seems to use GSL FFT whereas the Analysis>Forward FFT and Analysis>Inverse FFT use fft2D.cpp.

I have corrected an error in fft2D.cpp which stops a crash on Ubuntu 14.04 and compared results using both FFT algorithms and found them to match.

Will move to testing it on Windows.

[Dan Nixon]

Fix for inverse FFT on matrix

Refs #7586

Changeset: b4d404c496d4c8d77708030cc3b7b5f0d1b7ac38

[Dan Nixon]

Analysis>Inverse FFT now works on Ubuntu and I'm confident is should on other platforms. I could do with this being tested on both Windows 7 and OS X to confirm.

I haven't been able to test the issue with Analysis>FFT... causing crashes.

[Dan Nixon]

Note this needs to be tested on Ubuntu, Windows and OS X before being closed.

[Peter Peterson]

Should be on the current release

[Nick Draper]

Tested and works on Windows 7, Needs someone to test on the Mac now really

[Michael Reuter]

The FFT menu options seems to work with all options now on OSX and I check RHEL6 and Fedora 20 for good measure. However, the Forward FFT and Inverse FFT menu options no longer seem to do anything relevant on any platform (OSX, Windows, RHEL6, Fedora 20). Having a matrix and then invoking one of those options just clears the matrix. Not very helpful. This used to work on RHEL6. Should this be fixed on this ticket or should another one be generated?

[Dan Nixon]

I'm not able to reproduce that issue on Ubuntu 14.04, both the forward and inverse FFT menu options produce an FFT in the original matrix (unless you refer to the fact that the original matrix is overwritten?)

[Michael Reuter]

The original matrix is not overwritten but simply disappears and no results are generated.

[Dan Nixon]

Since the FFT menu seems to be fine now is the easiest option here to just remove the Forward and Reverse FFT menu options since they only seem to duplicate what can be done in the FFT menu?

[Dan Nixon]

I tried the "Forward FFT" and "Inverse FFT" options under Windows and they both seem to work fine for me.

[Dan Nixon]

Just realised what the problem is here, the matrix is only cleared when it contains empty cells.

A warning and doing nothing would be better in this case.

[Dan Nixon]

Merge branch 'master' into bugfix/7586_matrix_inverse_fft_fix

Refs #7586

Changeset: 8e3b7bc8a745d2cb630cd7d60d18afb1d33b54a6

[Dan Nixon]

Warn when martix has NaN values before FFT

Refs #7586

Changeset: 03073e3902588cf6d09b99407872229dabcc00ee

[Dan Nixon]

To test:

1. Create a matrix, resize it and add some data such that every cell has a value
2. Run the "Forward FFT..." and "Inverse FFT..." from the Analysis menu
3. Repeat step 1 but leave some cells blank
4. Run the "Forward FFT..." and "Inverse FFT..." from the Analysis menu, this should give a warning and leave the original data intact

[Dan Nixon]

Merge branch 'master' into bugfix/7586_matrix_inverse_fft_fix

Refs #7586

Full changeset: ​8e3b7bc8a745d2cb630cd7d60d18afb1d33b54a6

[Harry Jeffery]

Merge remote-tracking branch 'origin/bugfix/7586_matrix_inverse_fft_fix'

Full changeset: ​417a4b16c08dc53c87d2bb74ef5738997a0c9e6a

[Stuart Campbell]

This ticket has been transferred to github issue ​8431